Predicting Protein-Peptide Complex Structures by Accounting for Peptide Flexibility and the Physicochemical...

Area of Science:

• Computational Biology
• Structural Biology
• Biophysics

Background:

• Protein-peptide complex structure prediction is vital for understanding cellular mechanisms and developing peptide-based therapeutics.
• Peptide flexibility and accurate binding mode ranking present significant computational challenges in docking.
• Current methods struggle with the extensive conformational space of flexible peptides.

Purpose of the Study:

• To develop a novel computational strategy, MDockPeP2, for predicting protein-peptide complex structures.
• To address the challenges of peptide flexibility and binding mode ranking in protein-peptide docking.
• To provide a tool requiring only peptide sequence and protein crystal structure.

Main Methods:

• MDockPeP2 integrates physicochemical information from monomeric proteins with an exhaustive search strategy.
• The method employs a combination of global search and local flexible minimization.
• It was systematically assessed on a newly constructed database of 89 protein-peptide complexes.

Main Results:

• MDockPeP2 achieved success rates of 58.4% (top 10 models) for bound docking and 19.0% for unbound docking.
• On the peptiDB dataset, success rates were 62.0% (bound) and 35.9% (unbound) for top 10 models.
• For the LEADS-PEP dataset, MDockPeP2 achieved a 69.8% success rate (top 10 models).

Conclusions:

• MDockPeP2 effectively predicts protein-peptide complex structures, overcoming challenges posed by peptide flexibility.
• The method demonstrates high performance on diverse datasets, including those with longer peptides.
• MDockPeP2 offers a valuable computational tool for structural biology and drug discovery.